A first trial to assess the temperature dependence of the measurements:
This plot shows the magnetic flux vs. the ambient outdoor temperature. Dedicated temperature sensors directly at the fluxgate sensors will be installed soon.
Darrel Emerson did exhaustive tests on the charateristics, thermal dependencies and thermal and long-term effects. For determining
the temperature dependence of the magnetic field measurements, he used a 50-day
dataset, with 24-hr binning and a dynamical range of 5 deg C, concluding a
dependency of -46 nT/deg F = -82.8 nT/deg C. My very
naive first approach, based on a minute-binned correlation within 96 hours and a temperature range of 2 deg C,
results in a dependency of -160 nT/deg C.
The difference in coefficients may be due to the exact location where the temperature is measured. I will continue investigating this as I continue improving the insulation of the magnetometer.
Another interesting plot contains the raw measurements without compensating for offsets, as caused by passing cars, slamming fridge doors, operating the washing machine, getting skies out of the room where the magnetometer is measuring or just dropping a nail ;-)
During 2017-01-11 and 2017-01-29, the sensors were operated in an empty 40-liter cartboard box. The box was stored in a modestly temperatur stable storage. A DS18B20 temperature sensor (0.5oC accuracy, 0.0625oC readout) was placed inside the box. A 672-hr correlation with between the magnetic flux sensors and the temperature sensor:
On 2017-01-29, the cartboard box was filled with styrofoam pellets, with the temperature sensor immersed in the center of the box. A new temperatur measurement period has started.
From 2017-02-06 on, we experienced and interesting 15 deg Celsius temperature
drop of the outside temperature. The SAM thermometer followed, registering a
steady drop of about 2 deg Celsius. Since these measurements were accompanied by a
rather quite magnetic period, they were used to determine a preliminary
temperature dependency of
the sensors. From 2017-02-07 on, this dependency is experimentally corrected for in the data.
After an upgrade to a better insulated EPS box, a new calibration run was performed on March 5/6:
A first, 68-day correlation vs planetary Kp: